Re: [PATCH v5 1/2] selinux: add brief info to policydb

[William Roberts]

On Wed, May 17, 2017 at 8:24 AM, Sebastien Buisson
<sbuisson.ddn@xxxxxxxxx> wrote:
> 2017-05-17 17:09 GMT+02:00 William Roberts <bill.c.roberts@xxxxxxxxx>:
>> On Wed, May 17, 2017 at 7:59 AM, Sebastien Buisson
>> <sbuisson.ddn@xxxxxxxxx> wrote:
>>> 2017-05-16 22:40 GMT+02:00 Stephen Smalley <sds@xxxxxxxxxxxxx>:
>>>>> +     strcpy(*brief, policydb.policybrief);
>>>>> +     /* *len is the length of the output string */
>>>>> +     *len = policybrief_len - 1;
>>>>
>>>> Is there a particular reason to not just return policybrief_len here as
>>>> well, for consistency in the interface?  How do you intend to use this
>>>> value in the caller?
>>>
>>> As called in the other patch to expose policy brief via selinuxfs
>>> (sel_read_policybrief), the intent is to provide the caller with the
>>> length of the string returned.
>>> Or should I set *len to policy brief_len here, and just make the
>>> caller aware that the returned length is in fact the length of the
>>> buffer (i.e. including terminating NUL byte)?
>>
>> What is the caller supposed to do with length? This interface seemed kind of
>> odd. If it's guaranteed NUL byte terminated, do they even need length?
>
> The length is useful as an input parameter in case the caller provides
> its own buffer (instead of letting the function allocate one), and as

This is what I don't get, why doesn't the function just always allocate?

> an output parameter in case the buffer given in input is not large
> enough.

This interface seems "Windowsy" (inout parameters)... Iv'e been looking at
it on and off for a few days and it just seems odd. Not odd enough for me
to give it more negative review comments.

> In any case, it can spare the caller the effort of recomputing the
> length. As an example, sel_read_policybrief() in the 2/2 patch needs
> to know the length of the string to put in the user buffer.

Oh yeah, IIRC offhand, you're adding each LSMs brief info and using strcpy
+ length instead of strcat avoiding the null iteration?